Skip to main content
The Controlled Batch Runner is controlled readiness tooling for proving Eval Labs can complete the platform lifecycle end to end. Owner/admin and evaluator roles can use it; tester cannot.

​
What it is

The Controlled Batch Runner is internal platform-readiness infrastructure. Canonical route: 
/lucia/batch-runner
It runs controlled batches through the same lifecycle:
create runexecute Lucia promptssave reviewsfinalize runpersist to Supabaseverify Run Historyverify Global Analysiscompact localStorage

​
Supported checkpoints

The current product supports:
  • 1-run smoke
  • 3-run checkpoint
  • 10-run checkpoint
The 60-run readiness gate was completed by controlled batches.

​
What it proves

Controlled batch results can prove:
  • run creation works
  • Lucia response capture works
  • review generation works
  • review persistence works
  • run finalization works
  • Run History truth works
  • Global Analysis truth works
  • Supabase counts match UI truth
  • localStorage stays compact
  • user scoping does not visibly leak across the tested owner context
  • controlled batch lifecycle works end to end

​
What it does not prove

Controlled batch results do not prove:
  • Lucia is ready for real operator use
  • Lucia is human-approved
  • human evaluators agree with AI scoring
  • broad employee rollout is complete
  • every future access/security decision is complete

​
Product boundary

The Controlled Batch Runner is not the normal Auto-generated Prompt Tester. It should not be presented as the simple tester onboarding workflow. It is available to evaluator in the current role model, but it remains blocked for tester and unassigned users.